Barriers to the Employment of Welfare Recipients

1 downloads 52 Views 255KB Size Report
Lawrence. Aber, Harry Holzer, Rukmalie Jayakody, Sanders Korenman, James Kunz, Rebecca Maynard, Kara. Mikulich, Robert ... Nathaniel Anderson, Heidi.
Institute for Research on Poverty Discussion Paper no. 1193-99

Barriers to the Employment of Welfare Recipients

Sandra Danziger, Mary Corcoran, Sheldon Danziger, Colleen Heflin, Ariel Kalil, Judith Levine, Daniel Rosen, Kristin Seefeldt, Kristine Siefert, Richard Tolman Poverty Research and Training Center School of Social Work University of Michigan

June 1999

This research was supported in part by grants from the Charles Stewart Mott Foundation, the Joyce Foundation, the National Institute of Mental Health (R-24M551363) to the Social Work Research Development Center on Poverty, Risk, and Mental Health, and the Office of the Vice-President for Research at the University of Michigan to the Program on Poverty and Social Welfare Policy. Lawrence Aber, Harry Holzer, Rukmalie Jayakody, Sanders Korenman, James Kunz, Rebecca Maynard, Kara Mikulich, Robert Moffitt, Harold Pollack, Lauren Rich, Robert Schoeni, Alan Werner, Barbara Wolfe, Alan Yaffe, and James Ziliak offered helpful comments on a previous draft. Nathaniel Anderson, Heidi Goldberg, and Yunju Nam provided valuable research assistance. Barbara Ramsey typed the manuscript. We thank survey manager Bruce Medbery and the interviewing staff from the Institute for Social Research Survey Research Center and the Michigan Family Independence Agency, especially Charles Overbey, Steve Smucker, and Robert Syers. Any opinions expressed are those of the authors.

IRP publications (discussion papers, special reports, and the newsletter Focus) are now available on the Internet. The IRP Web site can be accessed at the following address: http://www.ssc.wisc.edu/irp/

Abstract

Dramatic reductions in welfare caseloads since passage of the Personal Responsibility and Work Opportunity Reconciliation Act of 1996 have not allayed policy concerns about the employability of recipients remaining on the rolls. Analysis of potential barriers to employment can address whether current recipients have problems that either singly or in combination make it difficult for them to comply with the new requirements for getting and keeping jobs. In this paper, we explore the prevalence and work effects of 14 potential barriers in a new survey of a representative sample of 753 urban single-mother recipients. We report the prevalence of the barriers and how their number predicts employment rates, controlling for demographic characteristics. We also analyze which individual barriers are associated with employment and how a model inclusive of a comprehensive array of barriers improves upon a traditional human capital model of the work effects of education and work and welfare history. Single mothers who received welfare in 1997 had higher rates of personal health and mental health problems, domestic violence, and children’s health problems than do women in national samples, but they were no more likely than the general population to be drug or alcohol dependent. Only 15 percent of respondents had none of the barriers and almost two-thirds had two or more barriers. The numbers of multiple barriers were strongly and negatively associated with working, and among the individual barriers, low education, lack of access to transportation, poor health, having drug dependence or a major depressive disorder, and several experiences of workplace discrimination reduced employment. Welfare-to-work programs need to be more finely targeted with respect to exemptions and service provision, and states should consider providing longer-term and enhanced supports for those who face low prospects of leaving welfare for employment.

Barriers to the Employment of Welfare Recipients

INTRODUCTION

The Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA) of 1996 ended the federal guarantee of cash assistance and replaced the Aid to Families with Dependent Children (AFDC) Program with the Temporary Assistance for Needy Families (TANF) Program. Individuals are limited to receiving TANF funds for 5 years, or less at state option.1 Such changes at the federal level reflect, in part, state-level experiments that had been conducted over the past two decades. Prior to 1996, more than half of the states had instituted work requirements for some portion of the welfare caseload (under the Job Opportunity and Basic Skills Program of the Family Support Act of 1988), and 31 states had received waivers from the federal government to test time-limited welfare receipt (U.S. Department of Health and Human Services, Administration for Children and Families, 1996). These state-level reforms, coupled with a strong economy, contributed to pre-PRWORA declines in the welfare caseload—between FY 1994 and FY 1996, the average monthly AFDC caseload dropped almost 14 percent (U.S. Department of Health and Human Services, 1996). With the implementation of TANF and the continuing robust economic recovery, caseloads have continued to decline—35 percent between August 1996 and June 1998 (U.S. Department of Health and Human Services, Administration for Children and Families, 1998; Ziliak et al., 1997). The caseload decline for Michigan from February 1996 to February 1997 (the month in which this study sample was drawn) was 15 percent. From August 1996 to September 1998, the number of recipients in Michigan fell by 39 percent, slightly above the rate for the U.S. These dramatic reductions have led policymakers, researchers, and advocates to analyze the employability of recipients remaining on the rolls and to evaluate what services might be required to foster their transition from welfare to work. Some have hypothesized that many personal problems—for example,

2 poor physical or mental health, lack of transportation, and/or low skills—diminish the labor market prospects of current recipients and may interfere with their ability to comply with expanded work requirements. Recipients with a complex set of such barriers, who are neither exempt nor provided specific help to resolve these problems, are especially vulnerable to losing assistance for failure to meet these requirements, even if they have no alternative means of support. Analysis of potential barriers to employment can reveal the extent to which current welfare recipients have problems that either singly or in combination interfere with participating in training programs, complying with new rules, and, ultimately, getting jobs, keeping jobs, and increasing wages. How much these potential impediments to work put women and children in jeopardy depends on what service programs, training programs, and employers do in response to the problems and whether they are addressed prior to the termination of the families from public assistance rolls. Currently, most state programs emphasize job search assistance services to move as many recipients as possible quickly into jobs. Typically they do not systematically assess whether undiagnosed barriers to employment—such as lack of basic work skills and experience, inadequate knowledge of workplace norms, transportation problems, health and mental health problems, substance abuse, and domestic violence—limit recipients’ capacities to work regularly (Seefeldt, Sandfort, and Danziger, 1998). As we suggest below, such a “work first” strategy may be appropriate for many welfare recipients who were on the caseload when PRWORA was passed. However, given the large decline in caseloads since its passage, recipients who have not yet entered the workforce are likely to have more of these problems than pre-1996 recipients. In this paper, we use a new survey of a representative sample of single mothers who were welfare recipients in an urban Michigan county to explore how such barriers, often ignored by previous welfare researchers and by policymakers, constrain employability. We answer four questions about these barriers:

3 •

How prevalent among welfare recipients is each of a large number of barriers to employment?



What percentage of recipients have multiple barriers?



Is the number of barriers associated with welfare mothers’ employment, and how much does employment decrease as the number of barriers increases?



Which individual barriers matter for employment and how much more do we learn when we examine a comprehensive set of barriers than when we predict welfare recipients’ employment as a function of their schooling, work experience, and past welfare receipt?

We begin with a review of the relevant literature that identifies a comprehensive set of potential barriers to the transition to work by welfare mothers. Then we describe our data, sample, measures, and methodology. We next present results showing that (1) welfare recipients in the sample have unusually high levels of some barriers to work, such as physical and mental health problems, domestic violence, and lack of transportation, but relatively low levels of other barriers, such as drug or alcohol dependence and lack of understanding of work norms; (2) recipients commonly have multiple barriers; (3) the number of barriers is strongly and negatively associated with employment status. In addition, we find that (4) the expanded model of barriers is a significantly better predictor of employment than is a model that includes variables traditionally measured, such as education, work experience, and welfare history. We conclude with a discussion of the implications of these results for understanding the employment of single mothers and for reforming welfare-to-work policies.

LITERATURE REVIEW

Most welfare-to-work programs now being operated by the states seek to move recipients into the workforce quickly. Typically, they do not conduct assessments for, or provide services to address, a wide array of potential employment barriers, even though previous studies indicate that a number of personal factors impede employment (for a more detailed review, see Kalil et al., 1998).

4 A sizable minority of recipients are unable to keep jobs and cycle between work and welfare (Harris, 1993, 1996; Pavetti, 1993; Spalter-Roth et al., 1995). Some recipients are unable to get jobs, while others get jobs only to lose them because of inadequate job skills (Bane and Ellwood, 1994; Harris, 1996; Wagner et al., 1998). Holzer (1996) surveyed 3,200 employers about entry-level jobs available to workers without a college degree and reported that most jobs required credentials (high school diploma, work experience, references) that many recipients did not have. For example, about half of all welfare recipients are high school dropouts, and about 40 percent have had no experience prior to their first welfare spell (Harris, 1996). Holzer (1996) also reported that most entry-level jobs required workers to perform one or more of the following tasks on a daily basis: reading and writing paragraphs, dealing with customers, doing arithmetic, and using computers. In contrast, the average welfare recipient reads on a sixth- to eighth-grade level and may not be able to perform many of these basic tasks (National Institute for Literacy, 1996). Another possibility is that recipients are not “work ready”—i.e., they do not understand or follow workplace norms or behaviors. Evaluations of the Project Match and New Chance Demonstrations report that many recipients lost their jobs because they failed to understand the importance of punctuality and the seriousness of absenteeism, and resented or misunderstood the lines of authority and responsibility in the workplace (Berg, Olson, and Conrad, 1991; Hershey and Pavetti, 1997). Perceptions of employer discrimination may also inhibit employment prospects. Employer audit studies demonstrate that African Americans and Latinos are less likely to receive job offers than are whites with comparable credentials (Turner, Fix, and Struyk, 1991), and qualitative data suggest that employers negatively stereotype African Americans (Kirschenman and Neckerman, 1991). Almost half of AfricanAmerican women in a Los Angeles survey report having experienced job-related discrimination (Bobo, 1995). Little information is available on whether employers negatively stereotype welfare recipients.

5 Mental health problems may further limit the employability of welfare recipients. High levels of depressive symptoms among recipients have been documented (Steffick, 1996; Olson and Pavetti,1996). In addition, many welfare mothers experience traumas—e.g., rape, domestic violence, sexual molestation—that put them at high risk for post-traumatic stress syndrome (PTSD). Among participants in a welfare-to-work program in New Jersey, 22 percent reported having been raped, 55 percent having experienced domestic abuse, and 20 percent having been sexually molested as a child (Curcio, 1996). Whereas previous studies document the negative consequences of mental health problems on men’s employment and work hours, little information is available on the nature of the relationship between such problems and work for welfare mothers (Jayakody, Danziger, and Pollack, 1998; Kessler and Frank, 1996). Substance abuse also might negatively affect employment. Estimates of the prevalence of substance abuse among welfare recipients range widely from 6.6 percent to 37.0 percent, depending in large part on the measure used (Olson and Pavetti, 1996). The 1992 National Household Survey of Drug Abuse reports that 15.5 percent of recipients were impaired by drugs or alcohol—twice the rate of nonrecipients. Olson and Pavetti (1996) hypothesize that mothers’ and children’s physical health problems could reduce employment. Rates of physical health problems are higher among welfare mothers and their children than among women and children in the general population (Loprest and Acs, 1995; Olson and Pavetti, 1996), and there is a positive association between women’s own employment and health (Kessler, Turner, and House, 1987; Bird and Freemont, 1991). Wolfe and Hill (1995) find that a single mother’s health affects her work effort through her potential wage rate and estimated value of public and private insurance. They further find that her child’s health affects a single mother’s number of hours of work, but not her probability of employment. Several evaluation studies of welfare programs suggest that health problems cause recipients to lose jobs (Hershey and Pavetti, 1997).

6 Involvement in violent personal relationships is another potential barrier to work. Domestic violence is present in the lives of a high percentage of women on welfare (for a review, see Raphael and Tolman, 1997 ). Bassuk, Browne, and Buckner (1996), Bassuk et al. (1996), and Lloyd and Taluc (1999) report lifetime prevalence rates of domestic violence ranging from 48 to 63 percent, and current rates of domestic violence ranging from 10 to 31 percent among women on welfare. Violent partners may sabotage mothers’ attempts to enter the workforce. This review of the literature suggests that we evaluate whether the employability of welfare recipients is reduced by such factors as low education, little work experience, a lack of the basic skills demanded by employers, a lack of knowledge about behavioral rules of the workplace, perceived experiences of workplace discrimination and harassment, physical and mental health problems, alcohol and drug dependence, and domestic violence. Olson and Pavetti (1996) note that the presence of any single problem may not be an insurmountable barrier to work, but the presence of multiple problems may reduce employment. Using data from the 1991 NLSY, they estimate that 30 percent of welfare recipients had more than one of the following problems: mother’s and child’s poor health, alcohol and drug dependence, depression, and low basic skills. However, expanding the definition of barriers to include milder forms of skill deficits and additional barriers, such as exposure to domestic violence, could substantially increase the percentage of the welfare population with multiple barriers (Olson and Pavetti, 1996, p. 27). In sum, many potential barriers to work, their prevalence and co-occurrence, and their effects on work have been ignored in past studies of welfare recipients and in policy discussions of how to move recipients from welfare to work. This study seeks to remedy these omissions.

7 DATA, SAMPLE, AND MEASURES

Data and Sample In late 1997, the Women’s Employment Study (WES) first surveyed a random sample of 753 single mothers with children and who were on the welfare rolls in an urban Michigan county in February 1997. The sample was systematically selected with equal probability from an ordered list of the universe of active single-mother cases of the Michigan Family Independence Agency. To be eligible for the sample, a woman had to be •

between the ages of 18 and 54;



white (non-Hispanic) or African American;2 and



a U.S. citizen.

The in-person interviews took about one hour to complete; the response rate was 86 percent. WES respondents were interviewed between September 1997 and December 1997, 7 to 10 months after the sample was drawn. Respondents were interviewed again in late 1998 and will be interviewed for a third time in late 1999 or early 2000. For a description of the sample and survey procedures, see Appendix C. In designing this study, we cast a wide interdisciplinary net over the potential problems that could be prevalent among welfare mothers and could impede their moving into the workforce and leaving the rolls. We included measures of traditional human capital variables, such as failure to complete high school and low work experience, but we extended our measures to focus on mental and physical health problems as well as other psychosocial and familial disadvantages.

Demographic Measures In late 1997, the time of the survey, 72 percent of the respondents continued to receive cash benefits. Of these welfare recipients, about half were fulfilling the state’s requirement to work at least 20 hours a week (see Appendix B for further information on Michigan’s welfare system changes since welfare

8 reform). Most nonworking welfare recipients (60 percent) had participated in the mandated job search training program within the year. Among the 28 percent of the respondents who were no longer receiving welfare, about three-quarters were working at least 20 hours a week, and about half were working at least 35 hours a week. Table 1 describes the employment status and demographic characteristics of the sample as a whole. Fifty-eight percent of all respondents were working at least 20 hours a week, as required as a condition of assistance in Michigan.3 Another 4 percent worked less than half-time. Most of the jobs were in the service sector and few provided benefits.4 Of those women who were employed at the time of the survey, almost half (30/62 percent) were working 35 hours or more a week. Fifty-six percent of respondents are African American and 44 percent are white. About 28 percent of the sample are between 18 and 24 years old; 46 percent are between 25 and 34, and 26 percent are between 35 and 54. Almost nine out of ten of the women lived in urban census tracts in the county. Although all respondents were single mothers in February 1997, 24 percent were living with a spouse or partner at the time of the survey. We do not know how many were cohabiting in February. About two-fifths were the primary caregivers for at least one child younger than 2 years, and about the same percentage for a child between the ages of 3 and 5. The average number of years of welfare receipt since turning age 18 was 7.3, with a range of 1 to 30 years.

Employment Barrier Measures Table 2 lists our measures of 14 barriers to employment.5 The cut-off points for measures of these potential barriers are defined as follows. Education, Work Experience, Job Skills, and Workplace Norms. A respondent is considered to have the education barrier if she neither graduated from high school nor received a GED.6

9 TABLE 1 Sample Characteristics Characteristic

Percentage of Respondents

Current welfare recipients

72%

Employment status Currently employed Working less than 20 hours/week Working 20–34 hours/week Working 35 hours or more

62% 4% 28% 30%

Race African American White

56% 44%

Age 18–24 years 25–34 years 35 years or more

28% 46% 26%

Residence Urban census tract Rural census tract

86% 14%

Marital status Living with spouse or partner at time of interview Other

24% 76%

Presence of young children Any 0–2 years Mean number of children 0–2 years

43% .49

Any 3–5 years Mean number of children 3–5 years

42% .51

Welfare history Mean number of years since age 18 in which received AFDC/FIP

7.3

Notes: The sample includes 753 women who received cash welfare in February 1997 and who were interviewed between September and December 1997. Because the respondents represent a random sample of all single-parent recipients in the county, no sample weights are utilized. For further information on the sample, see Appendix C.

10 TABLE 2 Measures of Employment Barriers Education, work experience, job skills, and workplace norms 1. 2. 3. 4.

Less than a high school education Low work experience (worked in fewer than 20 percent of years since age 18) Fewer than 4 job skills on a previous job (out of a possible 9) Knows 5 or fewer work norms (out of a possible 9)

Perceived discrimination 5.

Reports 4 or more instances of prior discrimination on the basis of race, gender, or welfare status (out of a possible 16)

Transportation 6.

Does not have access to a car and/or does not have a driver’s license

Psychiatric disorders and substance dependence within past year 7. 8. 9. 10. 11.

Major depressive disorder PTSD—post-traumatic stress disorder Generalized anxiety disorder Alcohol dependence Drug dependence

Physical health problems 12. 13.

Mother’s health problem (self-reported fair/poor health and age-specific physical limitation) Child health problem (has a health, learning, or emotional problem)

Domestic violence 14.

Severe abuse from a partner within past year

11 A respondent is considered to have low work experience if she worked in less than 20 percent of the years since she turned age 18. Respondents were asked about which of the following nine tasks they had performed on a daily or weekly basis in previous jobs: work with a computer; write letters or memos; watch gauges; talk with customers face to face; talk with customers on the phone; read instructions; fill out forms; do arithmetic; work with electronic machines. If a respondent had performed fewer than four tasks on a regular basis, she was classified as having a job-skills barrier. These skill questions were adapted from Holzer (1996) and Newman (1996). We asked respondents about the appropriateness of nine behavioral workplace norms. They answered “yes” or “no” to whether it would be a problem at work if they missed work without calling in, did not correct a problem pointed out by a supervisor, came to work late, made personal calls, argued with customers, left work early, took a longer break than scheduled, refused tasks not in the job description, and did not get along with a supervisor. Those who replied that five or more of these would not be a serious problem were classified as having this barrier. These questions were based on Berg, Olson, and Conrad (1991). Perceived Discrimination. Respondents were asked 16 questions about discrimination, including whether they had ever been refused a job, fired, or not promoted because of their race, sex, or welfare status (Turner, Fix, and Struyk, 1991; Bobo, 1995; Kirschenman and Neckerman, 1991).7 The women were asked if their current or most recent supervisor made racial slurs, insulting remarks about women, or insulting remarks about welfare recipients. They were asked about whether they had experienced discrimination because of race, gender, or welfare status on their current/most recent job. They were also asked whether they had been sexually harassed at work. Women who reported four or more instances of these experiences were classified as having this barrier. These questions were adapted from Bobo’s (1995) Los Angeles survey.

12 Transportation. We considered a respondent to have a transportation problem if she lacked access to a car and/or did not have a driver’s license. Mental Health and Substance Dependence. Mental health and substance dependence were assessed with diagnostic screening batteries for the 12-month prevalence of five psychiatric disorders in the Diagnostic and Statistical Manual, revised third edition [DSM-III-R] (American Psychiatric Association, 1987)—major depressive disorder, post-traumatic stress disorder (PTSD), generalized anxiety disorder, alcohol dependence, and drug dependence. Questions come from the Composite International Diagnostic Interview used in the National Comorbidity Survey (NCS), the first nationally representative survey to administer a structured psychiatric interview (Kessler et al., 1994). The items in each of the five indices are scored for clinical caseness, and all respondents who meet the scale criteria are defined as having the disorder barrier. Physical Health. Sample members were asked about physical limitations and to rate their general health using the health questions taken from the SF-36 Health Survey (Ware and Sherbourne, 1992). Respondents who rated their general health as poor or fair and who scored in the lowest age-specific quartile (based on national norms) of the multiple-item physical functioning scale were defined as having a health problem. Respondents who reported that at least one child had a physical, learning, or emotional problem that limited his/her activity were defined as having a child with a health problem. Domestic Violence. Domestic violence was assessed by the Conflict Tactics Scale, a widely used measure of family violence (Strauss and Gelles, 1986, 1990). We defined the barrier from the items indicating current (past 12 months) severe physical abuse. This subscale indicates whether the respondent has been hit with a fist or object, beaten, choked, threatened with a weapon, or forced into sexual activity against her will.

13 METHODOLOGY

The analysis sample is the 721 respondents who had no missing data on employment status or on any of the 14 barrier measures. We begin by estimating the prevalence of individual and multiple barriers in the sample. This tells us how many recipients face obstacles in these domains. Next, we examine whether the number of barriers a recipient has affects her employment status by estimating Equation 1, which expresses employment status as a function of the number of barriers, prior welfare receipt, and a series of demographic controls.

7

n

i =1

j =1

EMP = α 0 + ∑ α i N i + βW + ∑ θ j X j + µ1

(1)

where: EMP = 1 if working 20 or more hours/week; 0 otherwise Ni = 1 if the number of barriers = i; 0 otherwise; i = 1…6 N7 = 1 if the number of barriers is 7 or more; 0 otherwise W = number of years of prior welfare receipt Xj = set of demographic controls (marital status, race, age, number and ages of children, urban/rural residence) µ = random error term We estimate Equation 1 using logistic regression. As a last step, we investigate how each of the individual barrier indicators affects a recipient’s employment status by estimating Equation 2, which expresses employment status as a function of each of the 14 individual barrier measures, prior welfare receipt, and demographic controls. 14

n

k =1

j =1

EMP = β 0 + ∑ β k Bark + γW + ∑ θ j X j + µ 2

where Bark = a set of 14 dummy variables representing each of the barrier measures.

(2)

14 For comparison purposes, we also estimate a model that expresses employment status as a function of education, work experience, prior welfare receipt, and demographic controls—the model typically used in past analyses of welfare-to-work transitions. This comparison provides an estimate of how much an expanded set of barriers improves our ability to predict the employment of welfare recipients.

RESULTS

Prevalence of Specific Barriers Table 3 reports the prevalence among respondents of each of the 14 barriers in column l and (where possible) their prevalence in national samples of adult women in column 2. Recipients are much less likely to have graduated from high school and much more likely to have experienced transportation problems, mental health problems, physical health problems, child health problems, and severe physical abuse than women in the general population. On the positive side, recipients were no more likely to be drug or alcohol dependent than were adult women in the general population. About 30 percent of respondents had not finished high school (compared to 12.7 percent of women in this age range in the 1998 Current Population Survey), and about one-fifth had previously used fewer than four of the nine job skills. About one-tenth of recipients had little work experience, and a similar percentage knew five or fewer of the nine workplace norms. The finding that most recipients were familiar with work norms was a surprise, given that much of the job preparation training in “work first” programs assumes a general lack of this knowledge among recipients. About half of the recipients reported experiencing at least one instance of discrimination, and 13.9 percent reported four or more (out of 16) instances of these problems in their prior work experiences. Transportation problems were common—about half of the respondents lacked access to a car and/or did not have a license to drive. Comparatively few women nationally report lack of access to a vehicle in the household.

15 TABLE 3 Prevalence of Employment Barriers

Barriers

% in Sample with Barrier (1)

Education, experience, skills, and norms Less than high school education 30.1 Low work experience 10.2 Fewer than 4 job skills 21.1 Knows 5 or fewer work norms 9.1

% Women Nationally with Barrier (2)

12.7a

% in Sample Working 20+ Hours/Week with without Barriers Barriers (3) (4)

39.8* 46.1* 34.2* 56.7

65.4 59.0 64.0 57.8

46.7*

59.5

Perceived discrimination Reports 4 or more instances of discrimination

13.9

Transportation Has no car and/or driver’s license

47.3

7.6b

44.6*

69.4

Mental health and substance abuse Major depressive disorder PTSD Generalized anxiety disorder Alcohol dependence Drug dependence

26.7 14.6 7.3 2.7 3.3

12.9c

48.0* 55.0 54.5 70.0 40.0+

61.2 58.1 57.9 57.3 58.3

39.0*

62.2

Physical health Mother has health problem Child has health, learning, or emotional problems Domestic violence Severe abuse within last year a

4.3c 3.7c 1.9c

19.4 22.1

9.7d

48.5*

60.6

14.9

3.2–3.4

55.4

58.1

1998 Current Population Survey: % of all women aged 18–54 who do not have a high school diploma or equivalent. b 1990 Census: % of all women aged 18–54 who live in households with no vehicle available. c 1994 National Comorbidity Survey: % of all women ages 15–54 who meet criteria for clinical caseness on each of these disorders. d 1994 National Longitudinal Survey of Youth: % of all mothers aged 29–36 with children who have one of six limitations. e 1993 Commonwealth Fund Survey and 1985 National Family Violence Survey: % of all women aged 18 and over who report current severe physical abuse. +Difference between columns 3 and 4 is significant at the .10 level. *Difference between columns 3 and 4 is significant at the .05 level.

16 Mental health problems were common: 36 percent of respondents met the criteria for at least one of the five DSM-III-R diagnoses. More than a quarter had experienced a major depression within the past year, 15 percent met criteria for PTSD, and 7 percent met criteria for generalized anxiety disorder. These rates are considerably higher than those for women aged 15 to 54 in the NCS, where the rate of major depression was 13 percent and the rate of generalized anxiety disorder was 4 percent. There are no national estimates for 12-month prevalence of PTSD, but 29 percent of our sample met the criteria for a lifetime experience of PTSD, compared to less than 10 percent of women in the NCS. Self-reported substance dependence was low in this sample and comparable to prevalence rates in the NCS. Despite the popular view that many women on welfare abuse alcohol and drugs, only 3.3 percent of our sample met the criteria for drug dependence, and only 2.7 percent for alcohol dependence. It is possible that respondents (as well as the national sample of women) underreported their alcohol and drug dependence, because dependence involves a stricter definition of impairment than does use or abuse.8 These substance dependence rates are somewhat lower than those reported among welfare recipients in national samples (Jayakody, Danziger, and Pollack, 1998). About one in five mothers had a health problem, and a similar fraction reported that at least one of their children had a health, learning, or emotional problem. Our composite measure of maternal health is not directly comparable to measures used in national surveys (thus not reported in Table 3), but we can compare mothers’ scores on its two components to findings from national surveys. Respondents were twice as likely as the general population of adult women to report physical limitations and three to five times as likely to report their general health as poor or fair as are nonelderly women nationally (McDowell and Newell, 1996). The prevalence of having a child with an activity-limiting physical, emotional, or learning condition was twice as high as among an NLSY sample of young mothers. However, the measure in this national sample is slightly more complex, and the age range of the mothers is narrower than in our sample.

17 About 15 percent of the women reported being severely physically abused by a husband or partner in the last year. This rate is four to five times the national average (Straus and Gelles, 1986, 1990; Plichta, 1996) but similar to rates reported in other studies of welfare recipients (Raphael, 1995). We hypothesized that each of the 14 characteristics listed in Table 3 is a potential barrier to work. Bivariate analysis documents a strong relationship. The last two columns in the table show the proportion who work at least 20 hours a week, first for women with and then for those without each barrier. For nine of the 14 barriers, women who have the barrier are significantly less likely to work than those without the barrier. These barriers include (1) less than a high school education or GED, (2) little work experience, (3) previously used fewer than four of nine job skills, (4) had four or more prior perceived experiences of job discrimination, (5) lacked access to a car and/or a driver’s license, (6) major depressive disorder, (7) drug dependence, (8) poor health, and (9) had a child with health, learning, or emotional problems. For example, 34.2 percent of women with few job skills worked at least 20 hours a week, compared to 64.0 percent of those with more previous job skills. There were few differences between whites and African Americans in the prevalence of individual barriers. As shown in Appendix A, only two of the 14 individual barriers—lacking a car and/or driver’s license and having previously used fewer than four job skills—differ significantly between AfricanAmerican and white non-Hispanic respondents, and African-American single mothers are more likely to have these barriers. The distribution of the number of barriers did not differ significantly by race.

Prevalence of Multiple Barriers Most current recipients meet the criteria for several barriers, thus potentially compounding their disadvantages in the labor market. One or two barriers may have little effect on employment, but multiple barriers could seriously impede employment. For example, mental health and physical health problems might require frequent doctor visits, leading to absences from work. One of these problems alone might not

18 interfere with work, but in combination with low education and few job skills, it could create obstacles on the job or in job search. Lack of a high school diploma by itself does not constitute a rigid barrier to employment, but an employer might be less willing to hire a high school dropout who also lacks work skills, has transportation problems, and is depressed. Figure 1 reports the distribution of the number of barriers among respondents. Almost all recipients (85 percent) had at least one barrier to employment. In contrast to Olson and Pavetti’s 1991 NLSY data (1996), where most recipients had only one barrier, only 21 percent of our respondents had just one of the 14 barriers. Multiple barriers were common—37 percent had two or three barriers, 24 percent had four to six barriers, and 3 percent had seven or more barriers. Given the overall high prevalence of comorbidity of barriers across this wide range of domains, we next examine how the number of barriers is related to employment.

Barriers and Employment Outcomes We explored the association between the number of barriers and the respondent’s employment status at the time of the interview by estimating Equation 1. The dependent variable in Equation 1 indicates whether a woman was working at least 20 hours a week. Independent variables include seven dummies for the number of barriers, number of years received welfare since age 18, and demographic control variables (race, marital status, age, residence in an urban census tract, and whether she cares for young children). Model I in Table 4 reports the results. The probability that a woman worked at least 20 hours (as required for a recipient to be in compliance with Michigan’s welfare rules) decreases as her number of potential barriers to work increases. All of the coefficients on the number of barriers are significant and negative. The sizes of the coefficients cluster closely into five groups: 0 barriers, 1 barrier, 2–3 barriers, 4–6 barriers, and 7 or more barriers. We next estimated an equation that replaces the seven dummy variables in Model I with four dummy variables

FIGURE 1 Number of Barriers Experienced

6 4%

7 or more 3% 0 15%

5 8%

4 12% 1 21%

3 17%

2 20%

20 TABLE 4 Effects of Multiple Barriers on Whether Woman Works 20 or More Hours/Week Model II

Model I

Number of barriers 1 2 3 4 5 6 7 or more

Coefficient

S.E.

Odds Ratio

-0.506+ -0.844* -1.059* -1.862* -1.532* -2.087* -4.228*

0.292 0.289 0.299 0.329 0.371 0.478 1.055

0.604 0.430 0.347 0.155 0.216 0.124 0.015

Grouped barriers 1 2–3 4–6 7 or more Demographics Married/cohabits African American Urban census tract Age 25–34 35 and over Number of children 0–2 years old 3–5 years old Years on welfare Constant -2 log likelihood Number of observations + Significant at the .10 level. * Significant at the .05 level.

Coefficient

S.E.

Odds Ratio

-0.506+ -0.939* -1.790* -4.223*

0.292 0.265 0.287 1.055

0.603 0.391 0.167 0.015

-0.277 -0.080 0.514*

0.201 0.181 0.250

0.797 0.924 1.671

-0.208 -0.069 0.508*

0.201 0.181 0.250

0.812 0.933 1.662

0.565* 0.802*

0.222 0.314

1.759 2.230

0.580* 0.820*

0.222 0.312

1.786 2.269

-0.292* -0.041 -0.047* 1.025

0.149 0.126 0.020 0.363

0.747 0.959 0.954

-0.267+ -0.043 -0.047* 0.997

0.148 0.126 0.020 0.362

0.766 0.958 0.954

876.6 721

878.8 721

21 representing these five clusters. Model II in Table 4 reports the results. Again, employment decreases sharply as the number of co-occurring barriers increases, and all four of the coefficients are significant. Of the demographic characteristics, living in an urban census tract, mother’s age, number of years of prior welfare receipt, and number of very young children were significantly associated with women’s employment. Table 5 converts the estimated regression coefficients reported in Model II of Table 4 into predicted probabilities. The values represent the probabilities that a single, African-American mother, aged 25 to 34, who lived in an urban census tract, had one child under 2 years old, had no children between ages 3 and 5, and had received welfare for 7 years would work at least 20 hours a week if she had 0 barriers, 1 barrier, 2–3 barriers, 4–6 barriers, or 7 or more barriers. The results are striking—the greater the number of barriers, the less likely the woman is to work. Women with only one barrier were almost as likely to work as were women with no barriers (80.5 versus 71.3 percent). After that, employment drops sharply as the number of barriers rises. A woman has three chances in five of working if she has two or three barriers; two chances in five of working if she has four, five, or six barriers; and only a one chance in 20 of working if she has seven or more barriers to work.9 The coefficient on race in Table 4 is small and not significant, so the employment probabilities for whites, holding other characteristics constant, are similar to those shown in Table 5 for African-American single mothers. The probabilities for a white single mother with these characteristics with 0, 1, 2–3, 4–6, and 7+ barriers are 81.6, 72.7, 63.4, 42.5, and 6.1 percent, respectively.

Human Capital versus Expanded Barrier Model The analyses so far show that multiple barriers are associated with diminished employment among welfare recipients. However, the analyses do not identify which of the individual barriers have the greatest effects, and they do not show how well the expanded set of indicators improves our understanding over

22 TABLE 5 Employment Probabilities by Number of Barriers Number of Barriers

Probability of Working 20+ Hours/Week* (%)

0

80.5

1

71.3

2–3

61.7

4–6

40.8

7 or more

5.7

*Given that respondent is single, African American, lives in an urban census tract, is 25–34 years old, has one child aged 0–2, has no children aged 3–5, and has received welfare for 7 years. Predicted probabilities are based on the coefficients in Model II of Table 4.

23 previous studies that typically predict employment on the basis of recipients’ education, work experience, welfare experience, and demographic characteristics. Table 6 presents this analysis by estimating two versions of Equation 2. Model I reports results when employment status is regressed on schooling, work experience, years of welfare receipt, and demographic controls—the human capital model used in prior research based on measures typically available in data sets. The results are consistent with these studies. The presence and number of young children, more years on welfare, and lower levels of schooling are negatively associated with employment. Model II in Table 6 reports results when employment status is regressed on work experience, schooling, years of welfare receipt, our additional 12 barrier measures (many of which are not available in national data sets), and demographic controls.10 This expanded model is a better overall predictor of employment, and most of the barrier measures that had significant bivariate associations with employment (shown in Table 3) remain significant in the full model. In addition to education, six other barriers are negatively and significantly associated with working at least 20 hours a week: lack of access to transportation, few work skills, being in poor health, being drug dependent, being depressed, and perceiving four or more instances of workplace discrimination. In addition to these barriers, being 18–24 years old, having young children, not living in an urban census tract, and a higher number of years of prior welfare receipt reduce employment. Factors such as race, marital status, lack of knowledge of workplace norms, and current domestic abuse are not significantly associated with employment. Table 7 converts the seven significant estimated regression coefficients reported for Model II in Table 6 into probabilities and compares the difference in the likelihood of working with and without each of the seven barriers. The first column shows the prevalence of each barrier in the sample (as reported previously in Table 3) for comparative purposes. Row 1 of column 2 reports the probability that the typical woman in the sample (single, African American, lives in an urban census tract, has a child aged 0 to 2 but not one aged 3 to 5, and has received welfare 7 years) who has no barriers is working 20 or more hours a

24 TABLE 6 Effects of Individual Barriers on Whether Woman Works 20 or More Hours/Week Model II

Model I Coefficient

S.E.

Odds Ratio

Demographics Married/cohabits African American Urban census tract Aged 25–34 Aged 35 and over # of kids aged 0–2 # of kids aged 3–5 Years on welfare

-0.187 -0.124 0.428+ 0.416+ 0.540+ -0.280+ -0.038 -0.047*

0.195 0.177 0.244 0.218 0.307 0.143 0.122 0.020

0.829 0.883 1.534 1.516 1.716 0.756 0.963 0.954

-0.105 0.017 0.594* 0.420* 0.703* -0.321* -0.038 -0.036

0.210 0.193 0.260 0.235 0.336 0.154 0.131 0.021

0.900 1.017 1.811 1.522 2.019 0.725 0.963 0.965

Barriers Less than HS education Low work experience

-0.942* -0.314

0.176 0.267

0.390 0.731

-0.599* -0.014

0.197 0.292

0.549 0.986

-0.895* -0.010 -0.713* -0.762* -0.490* 0.075 0.264 0.869 -1.072* -0.703* -0.244 0.255 1.029

0.221 0.299 0.245 0.186 0.208 0.257 0.339 0.605 0.521 0.224 0.210 0.262 0.325

0.409 0.990 0.490 0.467 0.613 1.078 1.302 2.385 0.342 0.495 0.775 1.290

Other barriers from WES Fewer than 4 job skills Knows 5 or fewer work norms Perceived discrimination Transportation problem Major depressive disorder PSTD (12 months) General anxiety disorder Alcohol dependence Drug dependence Mother’s health problem Child health problem Domestic violence Constant 0.582 -2 log likelihood Cox & Snell R2 Chi square (df) Number of observations *Significant at the .05 level. +Significant at the .10 level.

0.296 921.9 0.077 57.8 (10) 721

Coefficient

S.E.

Odds Ratio

840.6 0.176 139.2 (22) 721

25 TABLE 7 Relative Effects of Individual Barriers on Whether Woman Works 20 or More Hours/Week

Prevalence (%)

Predicted Probability of Working 20+ Hours*

Difference in Probability with and without Barrier

None

15.5

81.6



Less than HS education

30.1

70.9

10.7

Fewer than 4 job skills

21.1

64.4

17.2

Perceived discrimination

13.9

68.5

13.1

Transportation problem

47.3

67.4

14.2

Major depressive disorder

26.7

73.1

8.5

Drug dependence

3.3

60.2

21.3

Mother’s health problem

19.4

68.7

12.9

Barriers

*Given that respondent is single, African American, lives in an urban census tract, is 25–34 years old, has one child aged 0–2 years old, has no children aged 3–5, and has received welfare for 7 years. Predicted probabilities are based on the coefficients in Model II of Table 6.

26 week. Rows 2 through 8 of column 2 report the probability that a typical woman with only the barrier listed in that row was working 20 or more hours a week. The numbers in rows 2 through 8 of column 3 report the difference between the probability of working for women with no barriers and that of working for women with only the single barrier in that row. For example, almost half the women have the transportation barrier, and there is a 14.2 percentage point difference in the probabilities of working between those with and without access to a car and a driver’s license. The largest individual effects are for few work skills and drug dependence—close to 20 percentage points in both cases (17.2 and 21.3, respectively). But 21 percent of the sample have few work skills, whereas only 3.3 percent are drug dependent.

SUMMARY

We began this paper with four questions: •

How prevalent among women who were previously welfare recipients is each of a large number of potential barriers to employment, such as health problems, mental health problems, few job skills, and inadequate knowledge of workplace norms?



What percentage of women have multiple barriers?



Is the number of barriers associated with welfare mothers’ employment, and how much does employment decrease as the number of barriers increases?



Which individual barriers matter for employment, and how much do we gain by adding this comprehensive set of factors to a model of employment?

Barriers to work are quite prevalent. Only 15 percent of the respondents had none of the 14 barriers analyzed. These respondents, all of whom received welfare in February 1997, had much higher rates of personal health problems and health problems among their children, mental health problems, and domestic violence experiences than do women in national samples. In addition, substantial percentages had not completed high school, possessed few job skills, reported multiple instances of perceived workplace discrimination, and lacked access to a car and/or a driver’s license. Some positive findings emerged with

27 regard to these barriers—most recipients knew most workplace norms, most had at least some past work experience, and recipients were no more likely to be drug or alcohol dependent than were women in the general population. Given the high prevalence of many of the individual barriers, it is not surprising that multiple barriers were common. Almost two-thirds of the women had two or more potential barriers to work, and over one-quarter had four or more. These barriers were strongly associated with women’s employment in late 1997. The more barriers a woman had, the less likely she was to be working. For example, only twofifths of women with four, five, or six barriers and one in 20 with seven or more barriers worked at least 20 hours a week. We expect that the women in this panel study who remain on welfare over the next few years will, like current long-term recipients, have greater numbers of barriers and hence an even more difficult time securing employment. Finally, the individual barriers that were significantly associated with not working at least 20 hours a week, controlling for a variety of other factors, include low levels of education, few work skills, lack of access to transportation, poor health, drug dependence, major depression, and experiences of perceived workplace discrimination.

POLICY IMPLICATIONS

The continuing strong economic recovery has contributed to caseload declines and increased employment of welfare recipients (see also Ziliak et al., 1997). The fact that over half the women in the sample were fulfilling the work requirement of at least 20 hours a week suggests that when the county unemployment rate was about 5.7 percent, many recipients could meet policy expectations. However, most of these women were working in low-wage, service-sector jobs with few benefits (data not shown).

28 For the group with few or no barriers to employment, the low wages and lack of health insurance in many of these jobs indicate a continuing need for policies that make work pay. Refundable child care credits at the federal level, a state earned income tax credit, and further promotion of newly available child health coverage may be the kinds of reforms that will promote well-being among those who can succeed in moving into this labor market. And, when the economy turns down, there will be a need for transitional jobs or special unemployment insurance provisions for those who are able to work but cannot find an employer to hire them (Holzer, 1998). Even with the current extent of job availability, however, the heterogeneity of the welfare caseload means that different strategies will be needed to move mothers from welfare to work. For example, a Project Match report documented a variety of pathways that characterize the routes women take from welfare to work (Wagner et al., 1998). These trajectories may well be a result of both the risk profiles of the women and their access to services focused on their particular problems. For the sizable minority of women in our sample who had none or only one barrier (most of whom were already working at least 20 hours a week), the emphasis on “work first” and job search assistance common to many state programs may meet their needs in today’s robust economy. But, for the recipients who have more of these barriers, welfare-to-work programs and services may need to be more finely targeted. Several policy and program design questions raised by these findings include whether exemptions from work or temporary exemptions should be expanded. Should the states be required to deliver needed services to help these families and to facilitate the transition off assistance? Finally, how will professional service-delivery programs in the communities adapt to supply the needed services to welfare recipients and the working poor? Our results on the association between specific barriers reveal that improvements in each of the following areas would maximize the chances of large numbers of recipients moving into the labor force: •

developing access to transportation

29 •

increasing specific types of job skills



promoting the health status of mothers



treating major depression

Lack of a high school degree and perceptions of workplace discrimination are also significantly and negatively associated with the probability of becoming employed, but current state programs are not designed to address these issues. Finally, reductions in drug dependence would also likely promote employment, although only a small proportion of the caseload is drug dependent. Most current recipients, however, have multiple barriers and may require several types of services. For the sizable minority of recipients with two or three barriers, about 60 percent were predicted to work 20 hours or more a week. Reducing the number of barriers they face by one or two could potentially increase their employment at relatively modest costs. However, the costs of risk reduction depend on the particular combination of barriers the women have and the availability and effectiveness of the services provided. Another sizable group of recipients had four to six barriers, and only 40 percent of them were predicted to be employed. Here more intensive interventions may be required. Some recipients may need to be temporarily exempted from work while receiving counseling, schooling, or health or mental health services. Finally, a very small percentage of recipients (3 percent) had seven or more of the 14 barriers we examined. Virtually none of these women were predicted to be employed, even though they were not classified by the state as disabled. Their multiple barriers make it unlikely than an employer will hire them or that they will be able to hold a job over the long run. Enhanced and possibly long-term services will be required for them, ranging from literacy and skills training to screening and treatment for depression, substance abuse, and domestic violence. In addition, many of these recipients may need to work in sheltered workshops or community service jobs before they can handle the demands of the workplace. We doubt that

30 the states are adequately prepared to serve this very disadvantaged group of recipients, who may, given current trends in the caseload, become an increasing share in coming years. At this point, they are candidates for the 20 percent of the caseload that can be exempted from TANF’s federal time limit. Our findings that physical health problems, mental health problems, and domestic abuse are common in today’s welfare population have implications for service delivery programs, as well as for employment. More, better, and/or more accessible health, mental health counseling, and social service programs, along with transportation services and skills-enhancing opportunities, could potentially improve the quality of life for welfare families, as well as further their transition from welfare to work.

31 APPENDIX A Prevalence and Frequency of Barriers by Race Prevalence of Barriers by Race Race Selected Barriers

Black

Less than high school education Low work experience Fewer than 4 job skills Knows 5 or fewer work norms Perceived discrimination Transportation problem Major depressive disorder PTSD (12 months) General anxiety disorder Alcohol dependence Drug dependence Mother’s health problem Child health problem Domestic violence

29.0 10.5 17.3 7.7 12.7 35.8 29.3 15.1 6.2 3.4 3.1 21.0 23.1 16.7

White 31.3 10.4 23.9* 10.1 14.5 56.1* 24.3 14.2 8.4 2.2 3.6 18.1 20.1 13.0

*Difference between columns is significant at the .05 level.

Frequency of Barriers by Race Race Number of Barriers

Black

0 1 2 3 4 5 6 7+

18.0 22.3 18.0 18.6 9.9 7.1 3.7 2.5

Note: Differences between columns are not statistically significant.

White 14.0 19.6 22.3 16.4 13.5 7.6 3.7 2.9

33 APPENDIX B The Welfare System in Michigan, 1997

In 1997, the women in our sample received cash assistance through Michigan’s TANF program, called the Family Independence Program, or FIP. Under FIP, an applicant applies for assistance through local offices of the state-run Family Independence Agency. As part of the eligibility process, she must attend an orientation to the Work First program, Michigan’s employment program for welfare recipients. Once an applicant attends the orientation and is otherwise eligible, she and her family begin receiving cash assistance. Compared with those of other states, benefit levels are fairly generous: for most of the caseload (benefits vary slightly by geographic area) a family of three with no other income receives a maximum of $459 a month. Only 13 states pay more.1 During 1997, the average monthly grant received by FIP clients was $401.2 To continue receiving benefits, clients must work part-time (defined in 1997 as 20 hours a week at a minimum wage job, increased to 25 hours a week in mid-1998) or continue their participation in the Work First program. The only recipients excused from these requirements are mothers with newborn children less than 12 weeks old, the disabled or those caring for a disabled family member, those over age 65 or under age 16, and heads of child-only cases (in which only the children, but not the caretakers, receive the grant). Local offices are allowed some discretion in granting temporary deferrals due to shortterm problems. In 1997, approximately 20 percent of the caseload was excused from work.3 Otherwise, recipients who are not working must participate in a Work First program. Local Work First programs offer a variety of services related to job search, but their primary aim is to move recipients quickly into employment. Work First is overseen by the Michigan Jobs Commission and is run locally by regional Michigan Works! Agencies with services contracted out to local providers. Once working, recipients may keep the first $200 and an additional 20 percent of earned income without it affecting their grants. Assuming no other income, a family of three can earn approximately $775 a month before the case closes. Recipients who work for 3 consecutive months and earn at least $350 a month receive cashed out food stamps. Over the course of 1997, the proportion of cases reporting earned income increased from 31 percent to 35 percent (with average earnings of $489), while the proportion receiving cashed out food stamps increased from 12.5 percent to just under 17 percent.4

1

Computed based on information in Gallagher et al. (1998). This rank is not a completely accurate reflection of benefit levels because a few states report combined TANF and food stamp benefits, and other states (notably Wisconsin) do not provide cash benefits to all persons on the caseload. 2

Calculations based on data from Michigan Family Independence Agency (1998).

3

The population from which our sample was drawn also excluded child-only and two-parent cases. We then excluded from the sample those deferred due to health reasons, but we included those who were temporarily deferred, including mothers caring for young children. Our reasoning was that in the time between the sample draw and the interview (7 months at a minimum), their deferral status might have changed. 4

Calculations based on data from Michigan Family Independence Agency (1998).

34 Those who do not comply with the work requirements face penalties. In the early part of 1997, clients determined as noncompliant had their benefits reduced by 25 percent, followed by case closure after 12 months of noncompliance. Beginning April 1997, the sanction policy changed so that cases close after 4 months of noncompliance. Additionally, new applicants are not eligible for assistance beyond 60 days if they fail to cooperate with program requirements. Over the course of 1997, very few cases were closed due to sanctions; approximately 2,000 cases (out of a caseload averaging 144,764) were closed for noncompliance with employment requirements. Overall, the state experienced a caseload decline of 14 percent in 1997 (from approximately 155,560 cases in January to 133,300 cases in December).5 Additionally, child care and medical assistance remain available to women while they receive cash assistance and after they leave the rolls.6 Child care is completely subsidized for cash assistance recipients who are working or participating in Work First. Working families not on the cash assistance rolls are eligible for a child care subsidy if their income falls below 85 percent of the state’s median income. These families are expected to pay a portion of the cost of care. During 1997 the child care caseload grew from approximately 34,600 cases with 59,000 children in January to 48,200 cases with 85,850 children in December.7 Medical assistance is provided to cash assistance recipients through the Medicaid program, and families who leave welfare due to increased earnings are eligible for up to 12 months of transitional Medicaid coverage. Children in low-income working families may also be eligible for coverage through MIChild, Michigan’s version of the Federal Child Health Insurance Program. Finally, unlike many states, Michigan has not instituted any state time limit for receipt of cash assistance. Although the 1996 federal legislation prohibits federal funds from being used for families for more than 60 months, state officials have indicated a willingness to continue support for families who are complying with program requirements but reach the federal time limit (these cases will begin to appear in October 2001). However, to date no program or provisions have been designed to serve this group.

5

All calculations in this paragraph are based on data reported in “Assistance Payments Statistics,” Michigan Family Independence Agency, 1997 volumes. 6

For more information on this topic, see Seefeldt et al. (1998).

7

Data reported in “Assistance Payments Statistics,” Michigan Family Independence Agency, January and December 1997 volumes.

35 APPENDIX C Sample Description and Survey Procedures The Women’s Employment Study is a simple random sample (n=753) systematically selected with equal probability from an ordered list of eligible women (n=8,875). To be eligible, women had to reside in this Michigan county and receive cash assistance in February 1997 and meet the following requirements:8 • • • •

single mothers with children US citizens between 18 and 54 years old racial identity of white or African American

To derive a representative sample of the metropolitan area and the population of these cases, staff at the Institute for Social Research Survey Research Center proportionally selected cases by ZIP code, race (white or African American), and age. Our response rate of 86.2 percent is calculated by dividing the interviewed cases by the sample cases (753/874). Excluded nonsample cases (n=26) include instances in which the sample person resided outside of the sample county, no housing unit existed at the address, or the sample person was institutionalized for the duration of the data collection period. We examined the correspondence between demographic distributions for the population and the final interviewed sample on the basis of race, age, months on welfare, number of people on case, employment codes, and monthly reported income. Based on these comparisons, the sample of interviewed women appears to be free of systematic bias. The one statistically significant difference—55.8 percent of interviewed sample is African American compared to 54.1 percent of universe—is not substantively significant. Once the sample was selected, letters of introduction were sent that included an 800 telephone number for respondents to call to arrange an interview. Institute for Social Research interviewers from the community conducted face-to-face interviews. Interviewers were instructed to complete domestic violence and life event history sections only if complete confidentiality could be assured. The average interview time was approximately 61 minutes. The average number of contacts to complete an interview was four; onequarter of the cases required six or more contacts. Respondents received $20 for completed interviews. There were no partial interviews. At the conclusion of the interview, respondents were given a resource list containing the names and telephone numbers of agencies and community organizations that offer a variety of emergency services and other resources.

8

We excluded noncitizens and other racial/ethnicity groups because both comprise a very small proportion of the population in the county.

36 APPENDIX C, continued Women’s Employment Study: Wave 1 Comparisons between Universe, Selected Sample, Interviewed Sample, And Non-Interviewed Sample by Selected Variables

Variable

N

Universe %

Total Selected Sample N %

Eligible Release Sample Interviewed Not Interviewed N % N %

Total

8,875

100.0

900

100.0

753

100.0

121

100.0

Race Black White

4,803 4,072

54.1 45.9

483 417

53.7 46.3

420 333

55.8 44.2

52 69

43.0 57.0

p=.0087 (group means: interviewed vs. not interviewed by race are significantly different using chi square test)

Age

8,875

900

753

121

Mean Median

29.1 27.9

29.2 28.0

29.3 28.0

28.7 28.0

High 75% quartile 25% quartile Low

55.0 33.9 23.3 16.2

54.1 34.1 23.2 16.7

54.1 34.3 23.5 17.7

51.2 32.6 23.0 16.7

p=.3890 (group means: interviewed vs. not interviewed by age are not significantly different using t-test)

Months on welfare Mean Median

8,681

High 75% quartile 25% quartile Low

878

736

117

32.4 17.0

32.0 17.0

31.8 17.0

35.9 16.0

297.0 42.0 6.0 1.0

280.0 41.0 6.0 1.0

280.0 42.0 6.0 1.0

213.0 43.0 6.0 1.0

p=.3130 (group means: interviewed vs. not interviewed by months on welfare are not significantly different using t-test)

# of people on case Mean Median High 75% quartile 25% quartile Low

8,875

900

753

121

4.3 4.0

4.4 4.0

4.4 4.0

4.1 3.0

23.0 5.0 3.0 2.0

25.0 5.0 3.0 2.0

23.0 5.0 3.0 2.0

14.0 5.0 3.0 2.0

p=.1021 (group means: interviewed vs. not interviewed by number of people on case are not significantly different using t-test)

37 Endnotes 1

To date, 20 states have adopted shorter time limits (National Governors’ Association, 1997). However, most of these states allow extensions and exemptions to the shorter limit. 2

Given the demographic composition of this urban county, we excluded about 3 percent of the cases where the single parent was not a citizen and was neither non-Hispanic white nor non-Hispanic black. In the 1990 census, only 2.1 percent of the population was Hispanic. For further information on the sample, see Appendix C. 3

The work requirement for remaining in compliance with welfare rules increased to 25 hours a week in 1998 (see Appendix B). 4

Average weekly earnings for workers in the sample were $208. Thirty-nine percent worked in the service sector and 41 percent worked in wholesale or retail trade. About 20 percent of those who were working and no longer receiving cash assistance had no health insurance. 5

Other measures included in the survey were not used in this analysis of the first wave data, including child care arrangements. We also collected information on child behavior concerns and parenting stresses, child care use and problems, access to social support, exposure to stressful life events and material hardship, perceived personal efficacy/mastery, residential mobility, and the like. Future project papers will analyze these data. Child care was an important barrier to employment in this sample, but our measure is confounded with the probability of working. More than two-fifths of the respondents reported that in the last year, they either lost or quit a job or were unable to take a job because of problems with child care or care of other family members. Those who reported this barrier were significantly less likely to be working 20 hours or more a week at the time of the interview than those who did not have this child care problem. However, we asked the questions in such a way that only those who participated in work or training in the first place could report that child care impeded their work prospects. Thus, we do not include this barrier in the set of barriers reported in this paper. The second wave of the survey includes a less endogenous measure of child care difficulties. 6

Seven percent of the sample has a GED. We treat them as high school graduates because they are quite similar to the high school graduates in our sample in terms of work experience, job skills, and extent of work. 7

These questions were adapted from surveys conducted by Lawrence Bobo, in Los Angeles (Bobo, 1995) and by James Jackson and David Williams in a Detroit-area study (Williams, 1996; Williams, Spencer, and Jackson, 1999). 8

Drug use is much more common. About one-fifth of the sample reported using an illegal substance at least once during the year prior to the survey. Most who used any drug used marijuana. For example, 16.2 percent reported use of marijuana/hashish, whereas only 2.5 percent used cocaine/crack. 9

The number of barriers is also correlated with continuing receipt of welfare (data not shown). Michigan’s income disregard ($200 per month plus 20 percent of additional earnings) allows many women who work part-time to continue to receive cash assistance, so many of those working part-time are still

38 receiving cash welfare. For example, of those with no barriers, 65 percent were still recipients, compared with 83 percent of those with six or more barriers (see also Appendix B). 10

The correlation matrix for the 14 barriers reveals no correlation above .33 and very few above .20.

39 References

American Psychiatric Association. 1987. Diagnostic and Statistical Manual of Mental Disorders, third edition, revised. Washington, DC: American Psychiatric Association. Bane, M. J., and D. T. Ellwood. 1994. Welfare Realities: From Rhetoric to Reform. Cambridge, MA: Harvard University Press. Bassuk, E. L., A. Browne, and J. C. Buckner. 1996. “Single Mothers and Welfare.” Scientific American 275 (4, October): 60–67. Bassuk, E. L., L. F. Weinreb, J. C. Buckner, A. Browne, A. Salomon, and S. S. Bassuk. 1996. “The Characteristics and Needs of Sheltered Homeless and Low-Income Housed Mothers.” Journal of the American Medical Association 276 (8, August 28): 640–646. Bird, C. E., and A. M. Fremont. 1991. “Gender, Time Use, and Health.” Journal of Health and Social Behavior 32: 114–129. Berg, L., L. Olson, and A. Conrad. 1991. “Causes and Implications of Rapid Job Loss among Participants in a Welfare-to-Work Program.” Paper presented at the Annual Research Conference of the Association for Public Policy and Management, Bethesda, MD. Bobo, L. 1995. “Surveying Racial Discrimination: Analyses from a Multiethnic Labor Market.” Working Paper #75, Russell Sage Foundation. Curcio, W. 1996. “The Passaic County Study of AFDC Recipients in a Welfare-to-Work Program: A Preliminary Analysis.” Passaic County, NJ: Passaic County Board of Social Services. Gallagher, L. J., M. Gallagher, K. Perese, S. Schreiber, and K. Watson. 1998. “One Year after Federal Welfare Reform: A Description of State Temporary Assistance for Needy Families (TANF) Decisions as of October 1997.” Urban Institute. Available on the Web at . Harris, K. M. 1993. “Work and Welfare among Single Mothers in Poverty.” American Journal of Sociology 99: 317–352. Harris, K. M. 1996. “Life after Welfare: Women, Work, and Repeat Dependency.” American Sociological Review 61: 407–426. Hershey, A. M., and L. A. Pavetti. 1997. “Turning Job Finders into Job Keepers: The Challenge of Sustaining Employment.” The Future of Children 7(1): 74–86. Available on the Web at . Holzer, H. 1996. What Employers Want: Job Prospects for Less-Educated Workers. New York: Russell Sage Foundation.

40 Holzer, H. 1998. “Will Employers Hire Welfare Recipients? Recent Survey Evidence from Michigan.” Discussion Paper no. 1177-98, Institute for Research on Poverty, University of Wisconsin–Madison. Available on the Web at . Jayakody, R., S. Danziger, and H. Pollack. 1998. “Welfare Reform, Substance Use and Mental Health.” Working paper, Michigan Program on Poverty and Social Welfare Policy, University of Michigan. Available on the Web at . Kalil, A., M. Corcoran, S. Danziger, R. Tolman, K. Seefeldt, D. Rosen, and Y. Nam. 1998. “Getting Jobs, Keeping Jobs, and Earning a Living Wage: Can Welfare Reform Work?” Discussion Paper no. 1170-98, Institute for Research on Poverty, University of Wisconsin–Madison. Available on the Web at . Kessler, R. C., and R. Frank. 1996. “The Impact of Psychiatric Disorders on Work Loss Days.” Working paper, University of Michigan. Kessler, R. C., K. A. McGonagle, S. Zhao, C. B. Nelson, M. Hughes, S. Eshleman, H.-U. Wittchen, and K. S. Kendler. 1994. “Lifetime and 12-Month Prevalence of DSM-III-R Psychiatric Disorders in the United States: Results from the National Comorbidity Survey.” Archives of General Psychiatry 51: 8–19. Kessler, R. C., J. B. Turner, and J. S. House. 1987. “Intervening Processes in the Relationship between Unemployment and Health.” Psychological Medicine 17: 949–961. Kirschenman, J., and K. M. Neckerman. 1991. “We’d Love to Hire Them But . . .: The Meaning of Race for Employers.” In The Urban Underclass, edited by C. Jencks and P. Peterson. Washington, DC: Brookings Institution. Lloyd, S., and N. Taluc. 1999. “The Effects of Male Violence on Women’s Employment.” Violence against Women 5: 370–392. Loprest, P., and G. Acs. 1995. “Profile of Disability among Families on AFDC.” Washington, DC: Urban Institute. McDowell, I., and C. Newell. 1996. Measuring Health: A Guide to Rating Scales and Questionnaires. New York: Oxford University Press. Michigan Family Independence Agency. 1998. “To Strengthen Michigan Families, Welfare Reform Data Monitoring, Data through November and December, 1997.” National Governors’ Association, Center for Best Practices. 1997. “Summary of Selected Elements of State Plans for TANF.” Washington, DC. National Institute for Literacy. 1996. Policy Update. August. Newman, K. 1996. Personal communication.

41 Olson, K., and L. Pavetti. 1996. “Personal and Family Challenges to the Successful Transition from Welfare to Work.” Washington, DC: Urban Institute. Pavetti, L. 1993. “The Dynamics of Welfare and Work: Exploring the Process by Which Women Work Their Way off Welfare.” Ph.D. dissertation, Harvard University. Plichta, S. B. 1996. “Violence and Abuse: Implications for Women’s Health.” In Women’s Health: The Commonwealth Fund Survey, edited by M. M. Falik. Baltimore: Johns Hopkins University Press, pp. 237–270. Raphael, J. 1995. “Domestic Violence: Telling the Untold Welfare-to-Work Story. Chicago: Taylor Institute. Raphael, J., and R. Tolman. 1997. “Trapped by Poverty, Trapped by Abuse: New Evidence Documenting the Relationship Between Domestic Violence and Welfare.” Chicago: Taylor Institute, and University of Michigan, Ann Arbor. Seefeldt, K., J. Sandfort, and S. K. Danziger. 1998. “Moving toward a Vision of Family Independence.” University of Michigan. Available on the Web at . Seefeldt, K. S., L. Pavetti, K. Maguire, and G. Kirby. 1998. “Income Support and Social Services for Low-Income People in Michigan.” Urban Institute. Available on the Web at . Spalter-Roth, R., B. Burr, H. Hartmann, and L. Shaw. 1995. Welfare That Works: The Working Lives of AFDC Recipients. Washington, DC: Institute for Women’s Policy Research. Steffick, D. 1996. “NLSY: Self-Esteem, Depression, and Wages.” Draft paper, University of Michigan. Straus, M. A., and R. J. Gelles. 1986. “Societal Change and Change in Family Violence from 1975 to 1985 as Revealed by Two National Surveys.” Journal of Marriage and the Family 48: 465–479. Straus, M. A., and R. J. Gelles, eds. 1990. Physical Violence in American Families: Risk Factors and Adaptations to Violence in 8,145 Families. New Brunswick, NJ: Transaction Books. Turner, M. A., M. Fix, and R. Struyk. 1991. “Opportunities Denied, Opportunities Diminished; Racial Discrimination in Hiring.” Washington, DC: Urban Institute. U.S. Department of Health and Human Services. 1996. Decline in AFDC Caseloads since FY1994. Washington, DC: USDHHS. U.S. Department of Health and Human Services, Administration for Children and Families. 1996. State Welfare Demonstrations. Washington, DC: USDHHS. U.S. Department of Health and Human Services, Administration for Children and Families. 1998. Change in Welfare Caseloads since Enactment of the New Welfare Law. Washington, DC: USDHHS.

42

Wagner, S., T. Herr, C. Chang, and D. Brooks. 1998. “Five Years of Welfare: Too Long? Too Short? Lessons from Project Match’s Longitudinal Tracking Data.” Chicago: Erikson Institute. Ware, J., and C. D. Sherbourne. 1992. “The MOS 36-Item Short-Form Health Survey (SF-36): Conceptual Framework and Item Selection.” Medical Care 30: 473–483. Williams, D. R. 1996. Personal communication. Williams, D. R., M. S. Spencer, and J. S. Jackson. 1999. “Race, Stress, and Physical Health.” In Self, Social Identity, and Physical Health, edited by R. J. Contrada and R. D. Ashmore. New York: Oxford University Press, pp. 71–100. Wolfe, B., and S. Hill. 1995. “The Effect of Health on the Work Effort of Single Mothers.” Journal of Human Resources 30: 42–62. Ziliak, J. P., D. N. Figlio, E. E. Davis, and L. S. Connolly. 1997. “Accounting for the Decline in AFDC Caseloads: Welfare Reform or Economic Growth?” Discussion Paper no. 1151-97, Institute for Research on Poverty, University of Wisconsin–Madison. Available on the Web at .